Open Research OnlineThe Open University’s repository of research publicationsand other research outputs

A spectroscopic study of IRAS F10214 + 4724Journal ItemHow to cite:

Serjeant, Stephen; Rawlings, Steve; Lacy, Mark; McMahon, Richard G.; Lawrence, Andy; Rowan-Robinson,Michael and Mountain, Matt (1998). A spectroscopic study of IRAS F10214 + 4724. Monthly Notices of the RoyalAstronomical Society, 298(2) pp. 321–331.

For guidance on citations see FAQs.

c© 1998 Royal Astronomical Society

Version: Version of Record

Link(s) to article on publisher’s website:http://dx.doi.org/doi:10.1046/j.1365-8711.1998.01522.xhttp://dx.doi.org/10.1046/j.1365-8711.1998.01522.x

Copyright and Moral Rights for the articles on this site are retained by the individual authors and/or other copyrightowners. For more information on Open Research Online’s data policy on reuse of materials please consult the policiespage.

oro.open.ac.uk

A spectroscopic study of IRAS F10214 þ 4724

Stephen Serjeant,1;2 Steve Rawlings,2 Mark Lacy,2 Richard G. McMahon,3 AndyLawrence,4 Michael Rowan-Robinson1 and Matt Mountain5

1Astrophysics Group, Imperial College London, Blackett Laboratory, Prince Consort Road, London SW7 2BZ2Astrophysics, Department of Physics, Keble Road, Oxford OX1 3RH3Institute of Astronomy, The Observatories, Madingley Road, Cambridge CB3 0HA4Institute for Astronomy, University of Edinburgh, Royal Observatory, Blackford Hill, Edinburgh EH9 3HJ5Gemini 8-m Telescopes Project, 950 N. Cherry Avenue, Tucson, AZ 85726, USA

Accepted 1998 February 11. Received 1998 February 11; in original form 1996 July 9

A B S T R A C TThe z ¼ 2:286 IRAS galaxy F10214 þ 4724 remains one of the most luminous galaxies in theUniverse, despite its gravitational lens magnification. We present optical and near-infraredspectra of F10214 þ 4724, with clear evidence for three distinct components: lines of width,1000 km s¹1 from a Seyfert 2 nucleus; & 200 km s¹1 lines which are likely to be associatedwith star formation; and a broad (,4000 km s¹1) C III] 1909-A emission line which isblueshifted by ,1000 km s¹1 with respect to the Seyfert 2 lines. Our study of the Seyfert 2component leads to several new results. (i) From the double-peaked structure in the Lya line,and the lack of Lyb, we argue that the Lya photons have emerged through a neutral column ofNH , 2:5 × 1025 m¹2, possibly located within the AGN narrow-line region, as proposed forseveral high-redshift radio galaxies. (ii) The resonant O VI 1032, 1036-A doublet (previouslyidentified as Lyb) is in an optically thick (1:1) ratio. At face value this implies an extremedensity (ne , 1017 m¹3) more typical of broad-line region clouds. However, we attribute thisinstead to the damping wings of Lyb from the resonant absorption. (iii) A tentative detection ofHe II 1086 suggests little extinction in the rest frame ultraviolet.

Key words: galaxies: active – galaxies: formation – galaxies: individual: FSC 10214 þ 4724– galaxies: starburst – gravitational lensing – infrared: galaxies.

1 I N T RO D U C T I O N

The z ¼ 2:286 IRAS galaxy FSC 10214 þ 4724 (hereafterF10214 þ 4724) is one of the most apparently luminous objectsin the Universe, and its discovery (Rowan-Robinson et al. 1991) ledto much speculation about its possible status as a protogalaxy. Thisspeculation was based on the extreme bolometric luminosity of theobject (Rowan-Robinson et al. 1991), and more specifically on thehuge gas mass and star formation rate inferred from the submmmolecular line and continuum detections (e.g. Solomon, Downes &Radford 1992; Rowan-Robinson et al. 1993).

This speculation was dampened by a series of papers whichproved that F10214 þ 4724 is being gravitationally lensed and, atall wavebands, is intrinsically in order of magnitude or moredimmer than it first appeared. Although a lensing bias wassuspected by a number of authors (e.g. Elston et al. 1994; Trentham1995), the first direct empirical evidence of strong gravitationallensing was provided by a deep near-infrared image (Matthews etal. 1994), which revealed an arc-like structure centred on a galaxyclose to the line of sight to F10214 þ 4724. Several sets of authorspublished lensing interpretations (Broadhurst & Lehar 1995;Graham & Liu 1995; Serjeant et al. 1995), which were confirmed

by the appearance of the HST image obtained by Eisenhardt et al.(1996): this image contained highly elliptical, high-surface-brightness features characteristic of strong lensing, as well as aclear counterimage. These papers also attempted to constrain theredshift of the system responsible for the gravitational lensing, ourcontribution (Serjeant et al. 1995) being spectroscopy of twogalaxies projected <1 and <3 arcsec from F10214 þ 4724. Thiswork revealed tentative 4000-A breaks at z . 0:90 in both galaxies,later confirmed using fundamental plane arguments in HSTimaging (Eisenhardt et al. 1996), and also tentatively supportedby a weak absorption line at z ¼ 0:893 in the F10214 þ 4724spectroscopy of Goodrich et al. (1996). The HST R-band imageof F10214 þ 4724 implies magnifications of ,100 (Eisenhardt etal. 1996) in this waveband, but it now seems likely that differentialflux magnification causes lower magnification factors for the moreextended structure, as argued by several authors.

Gravitational lensing appeared to offer a compelling explanationfor the extreme luminosity of F10214 þ 4724 (e.g. Broadhurst &Lehar 1995). As a result, the IRAS galaxy is no longer so extreme inits properties: indeed, in many respects it resembles localultraluminous infrared galaxies and Seyfert 2 galaxies. Never-theless, both Downes, Solomon & Radford (1995) and Green &

Mon. Not. R. Astron. Soc. 298, 321–331 (1998)

q 1998 RAS

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

Rowan-Robinson (1996) argue for a bolometric magnificationfactor of & 10 using arguments based on minimum blackbodysizes. F10214 þ 4724 remains one of the most intrinsically lumi-nous objects in the Universe. The importance in this object still liesin studying whether high-z hyperluminous activity, such as that seenin F10214 þ 4724, differs in all but scale from local objects, and indetermining the relative contributions of the starburst and AGNcomponents (Lawrence et al. 1993, 1994; Elston et al. 1994; Soiferet al. 1995; Goodrich et al. 1996; Kroker et al. 1996; Hughes,Dunlop & Rawlings 1997).

Prior to making the observations reported in this paper there hadbeen no direct evidence for the presence of an embedded broad-line(e.g., Seyfert 1 or quasar) nucleus in F10214 þ 4724, although high(<20 per cent) rest frame ultraviolet (UV) polarization (Lawrenceet al. 1993; Jannuzi et al. 1994) suggested that one was present. Thissituation changed with the deep spectropolarimetry of Goodrich etal. (1996) showing clear broad lines in polarized light. Thisextended the close spectral similarities between F10214 þ 4724and Seyfert 2 galaxies, specifically NGC 1068, which was firstremarked on by Elston et al. (1994).

Prior to our observations there had also been no reporteddetection of the narrow (& 200 km s¹1) emission lines expectedfrom any star-forming activity in F10214 þ 4724. A star-formingcomponent is expected if the analogy with NGC 1068 is to becomplete. This situation also changed during the preparation of thispaper. Using a novel imaging near-infrared spectrometer, Krokeret al. (1996) presented evidence for spatially extended narrowHa emission, just as expected if the Seyfert 2 nucleus ofF10214 þ 4724 is accompanied by a circumnuclear starburst.

In this paper we present, analyse and interpret optical and near-infrared spectroscopy of F10214 þ 4724. The details of data acqui-sition and analysis are given in Section 2. In Section 3 we compareour results with previous and contemporaneous spectroscopicstudies of F10214 þ 4724. In Section 4 we interpret the data onthe Seyfert 2 emission-line region, including a discussion of opticaldepth effects on the resonance lines, and some modelling of thespectrum using the photoionization code CLOUDY (e.g. Ferland 1993,1996). In this section we reach conclusions about the Seyfert 2properties of F10214 þ 4724 which differ significantly from thosereached by previous studies. In Section 5 we interpret the data on

322 S. Serjeant et al.

q 1998 RAS, MNRAS 298, 321–331

Figure 1. WHT spectrum of F10214 þ 4724. The spectrum was extracted from a full-width, zero-intensity aperture, thereby ensuring accurate spectro-photometry. This also included some light from a nearby companion object, particularly redward of <7500 A, as discussed by Serjeant et al. (1995). Theemission lines are marked with their identifications.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

the region responsible for the narrow (& 200 km s¹1) Ha line seenin our near-infrared spectrum: this feature is likely to be a signatureof star formation. In Section 6 we make some concluding remarkson the nature of F10214 þ 4724. Further data on galaxies fore-ground to F10214 þ 4724 are given in Appendix A, and ourattempts at identifying the weak emission line at 2067 A arediscussed in Appendix B.

2 DATA AC Q U I S I T I O N A N D A N A LY S I S

2.1 Optical spectroscopy

We observed F10214 þ 4724 on the nights of 1995 January 28 and30 with the ISIS spectrograph on the William Herschel Telescope(WHT), taking advantage of 0.7-arcsec seeing. We refer the readerto Serjeant et al. (1995) for full details of these observations. Thespectrophotometry is in good agreement with the continuummeasurements tabulated by Rowan-Robinson et al. (1993) as wellas (accounting for slit losses) the line parameters of Goodrich et al.(1996). There are significant disagreements between our measure-ments and those tabulated by Elston et al. (1994) – their line fluxesare typically larger by a factor of ,2 – but these disagreementscan plausibly be attributed to imperfect correction for the non-photometric observing conditions experienced by Elston et al.Neither the continuum nor any of the emission lines are resolvedspatially (with <0:35-arcsec pixels); limits on the extended lineemission are discussed by Serjeant et al. (1995).

2.2 Near-infrared spectroscopy

We observed F10214 þ 4724 on two occasions with the short (150-mm) camera and 3-arcsec (single pixel) slit of the 62 × 58 InSbCGS4 array (Mountain et al. 1990) on the UKIRT. The slit was

aligned at a position angle of 908. The first of these observations, on1992 February 5, used the 75 line mm¹1 grating to obtain a first-order spectrum centred near 1:6 mm; a series of 30-s exposures werearranged in sets of four, shifted by 0, 0.5, 1 and 1.5 in wavelength –this provided Nyquist sampling and ensured that a given wave-length was sampled by two pixels. The standard ‘ABBA’ noddingpattern with a nod of 24 arcsec (eight detector rows) was used. Thetotal exposure time was 9 × 4 × 2 × 30 s for each point in the 126-pixel spectrum. The second observation, on 1992 March 23,employed the 150 line mm¹1 grating to obtain a second-orderspectrum centred near 2:16 mm. The nod in this case was 30 arcsec(10 rows), and the total exposure time 10 × 4 × 2 × 40 s per point.Spectra were wavelength-calibrated using xenon/argon lamps andnight-sky hydroxyl lines, flux-calibrated using HD 105601, and thepositive and negative channels were extracted from their respective3-arcsec-wide rows. The wavelength calibration is accurate to,0:003 mm in H and 0:001 mm in K, and the spectral resolvingpowers were 250 at H and 1400 at K.

3 R E S U LT S A N D C O M PA R I S O N W I T H OT H E RS P E C T RO S C O P Y

3.1 Optical spectroscopy

Fig. 1 shows our WHT optical spectrum of F10214 þ 4724; linefluxes, widths and redshifts are tabulated in Table 1. Deeper spectracovering the region above 3900 A have been presented by Soifer etal. (1995) and Goodrich et al. (1996), and an MMT spectrum,comparable in sensitivity and wavelength range to our data, ispresented by Close et al. (1995); see also Rowan-Robinson et al.(1991) for the discovery spectrum. The velocity profiles of all theemission lines except Lya are, with FWHMs around 1000 km s¹1,extremely similar. These linewidths are small compared to quasar

A spectroscopic study of IRAS F10214 þ 4724 323

q 1998 RAS, MNRAS 298, 321–331

Table 1. Measurements from the WHT spectrum of F10214 þ 4724.

Notes: Errors on the line fluxes represent ,90 per cent confidence intervals expressed as a percentage of the best line-flux estimate; for the strongest lines theseare dominated by roughly equal contributions from uncertainties in fixing the local continuum level, and in the absolute flux calibration. Linewidths wereestimated from the FWHM of the best Gaussian fit to each line; for the restframe UV lines, the range lower value assumes that the line-emitting region fills the 2-arcsec WHT slit, and the higher value that is broadened only by the seeing.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

broad lines, but large compared to AGN narrow lines, which aretypically a few hundred km s¹1 (e.g. Nelson & Whittle 1996).

The accuracy of our wavelength calibration means that theidentification of the bright doublet in the far-blue with O VI issecure, and we are forced to conclude that Close et al. (1995)were mistaken in preferring Lyb as the identification for thisfeature. This is probably the most clearly resolved example of theO VI doublet (see, e.g., Kriss et al. 1992a,b and Laor et al. 1994), afactor we will exploit in Section 4. Above 3900 A, comparison withthe published spectra shows that all the labelled feature are real. Theidentification of the (rest frame) 2470-A feature, which is also seenin NGC 1068, is new, but probably uncontroversial. The identifica-tion of the line at <1805 A is more uncertain: we have followedLacy & Rawlings (1994) by identifying it with Mg VI, a line presentin the CLOUDY models discussed in Section 4.2. A strong line at thiswavelength is also seen in NGC 1068 (Snijders, Netzer & Boksen-berg 1986). We prefer the Mg VI identification to either the Si II

1814.0-A multiplet or the [Ne III] 1814.6-A line: these weresuggested as possible identifications by Snidjers et al., and adoptedfor F10214 þ 4724 by Soifer et al. (1995) and Close et al. (1995).Note that the ionization potential of Mg5þ is 141.2 eV, and is thuseven more extreme than the 113.9 eV required to produce O5þ.

This leaves one unidentified line at a rest frame wavelength of2067 A. This line is also seen in the HST spectrum of NGC 1068(Antonucci, Hurt & Miller 1994), where it lies in a region confusedby underlying Fe II multiplets (see also Wills, Netzer & Wills 1980for a discussion of a 2080-A feature in the spectra of quasars), but inF10214 þ 4724 it is clearly a distinct narrow feature. Our attemptsat identifying this line are discussed in Appendix B.

The broad asymmetric base to the C III]1909 line is shown in moredetail in Fig. 3. This broad base is independently present (thoughmarginally) in the spectra taken on both nights. The data arereproduced by the sum of two Gaussian components with a broad(,4000 km s¹1) component, blueshifted by about 1000 km s¹1

relative to a component with a similar FWHM (i.e., <1000 km s¹1)to the other optical lines. Although lines of Si III] are expected to bepresent at some level in the blue wing of C III]1909 (see Fig. 2), wewere unable to obtain as good a fit using a superposition of <1000km s¹1 lines (Fig. 2). This implies that the previous detection of hintof a broad C III]1909 line is integrated light (Goodrich et al. 1996) isnot attributable solely to the neighbouring narrow emission lines.

There is also an apparent blue wing to the C II]2326 line, but thisfeature is sensitive to the correction for the atmospheric A band.The lack of broad components to other lines may be explainable bya combination of scattering albedo and reddening towards thescattering surface [or between the broad-line region (BLR) andthe scatterers, or within the scattering region itself]; also, lowersignal-to-noise ratios may mask broad features in the Lya and N V

region (Fig. 4).The Lya line has a number of unusual properties which have been

noted by other authors. First, it is weak relative to the otherultraviolet emission lines, e.g., N V. This appears to reflect anom-alously weak Lya emission: ignoring Lya, both NGC 1068 (e.g.Snidjers et al. 1986; Kriss et al. 1992b) and the IRAS-detectedSeyfert 2 galaxy NGC 3393 (Diaz, Prieto & Wamsteker 1988) haveultraviolet spectra similar to that of F10214 þ 4724. Secondly, theLya line is double-peaked; this has been interpreted as requiring‘self-absorption’ (Soifer et al. 1995; Close et al. 1995). We willdiscuss these points further in Section 4.

324 S. Serjeant et al.

q 1998 RAS, MNRAS 298, 321–331

Figure 2. C III] line modelled as a narrow line and narrow Si III doublet. Theindividual model components are plotted as dotted lines, and the sum as afull line. Clearly, this makes an extremely poor fit to the data. The positionsof the Al III emission line and the z ¼ 1:361 Mg II absorber are also marked.

Figure 4. Expanded UV spectrum of F10214 þ 4724. Positions of majoremission lines are marked; also shown are the predicted positions of Lyb

associated with the observed Lya.

Figure 3. C III] line modelled as two Gaussians.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

The continuum depression shortward of Lya is within the range,though at the upper limit observed in quasars at similar redshifts(e.g. Warren, Hewitt & Osmer 1994), consistent with the expectedlack of a QSO proximity effect (e.g. Giallongo et al. 1996, andreferences therein). The He II 1086 line may be marginally detected.We confirm the presence of a Mg II 2800 absorption doublet atz ¼ 1:316 (see Fig. 2), but have insufficient sensitivity to confirmthe weaker z ¼ 0:892 doublet tentatively identified by Goodrichet al. (1996).

3.2 Near-infrared spectroscopy

Our H-band spectrum of F10214 þ 4724 is shown in Fig. 5, and ourK-band spectrum in Fig. 6. Other near-inrared spectroscopy ofF10214 þ 4724 has been published by Elston et al. (1994), Iwa-muro et al. (1995), Soifer et al. (1995) and Kroker et al. (1996).

We obtain an approximate lower limit on the [O III]4959 þ 5007to Hb ratio of 10. Iwamuro et al. (1995) claim a tentative detectionof Hb with an O III to Hb ratio of 27, although the Hb is a factor of 3brighter than the upper limit quoted by Elston et al. (1994).Iwamuro et al. used the OH airglow suppressor spectrograph onthe University of Hawaii 2.2-m telescope, so their detection shouldbe less prone to difficulties in sky subtraction. However, it is notclear how imperfect sky subtraction could lead to an underestimateof the Hb flux without a corresponding null detection of the[O III]4959, 5007-A doublet, since the sky lines are if anythingstronger near the [O III] lines, and the relative variations of the skylines are typically #10 per cent (Ramsay, Mountain & Geballe1992); there are also no strong atmospheric absorption lines in thevicinity, and the spatial variation of the sky spectrum is alsoexpected to be negligible. Nevertheless, the lower resolution H-band spectrum of Soifer et al. (1995) has a marginally detected bluewing to the [O III] lines, which may be Hb.

Weak features near the predicted position of the He II 4686 line

appears to be present, which might contribute to the unusual widthof the apparent 4686-A line in Soifer et al. (1995). However, none ofthese features is reliably detected, and none lies on any of thepositions of z ¼ 2:286 emission lines expected to be prominent(see, e.g., Section 4.4 below), which casts doubt on the reliability ofthe apparent detections. Soifer et al. (1995) quote a detection ofHe II 4686 A at an O III/He II ratio of 23, much fainter than the limit(,5) placed in Fig. 5; this detection is confirmed, albeit tentatively,by Iwamuro et al. (1995).

The Ha line profile in our CGS4 K-band spectrum (Fig. 6) isclearly resolved into a broader resolved component, with widthcomparable to the UV emission lines, and a narrower unresolvedcomponent of Ha and [N II]. We fit the K-band spectrum using thestsdas. fitting.ngauss package with the background level as a freeparameter, and with the following four Gaussians: first, threeunresolved (FWHM 200 km s¹1) Gaussians to model the narrowerHa line and satellite [N II] 6548, 6584-A lines, with the 6548-Acomponent constrained to have the same redshift and 1/3 the flux ofthe 6584-A component; second, a further Gaussian with uncon-strained amplitude, position and width. Although the broadercomponent is a blend of the broader Ha and satellite [N II] lines,we model it as a single Gaussian for simplicity.

The results, displayed in Fig. 7, were very similar to a deblendingwith three Gaussians of unconstrained position and width, andfurther deblending models with alternative simplifying assump-tions were attempted. In particular, the region between the narrowerHa and the narrower [N II]6584 lines was very poorly fitted inmodels without a broad excess, as might be expected from Fig. 7.The flux in this region is clear evidence for an additional compo-nent. From the variations in the narrower line flux between themodels we estimate the flux errors in both the narrower Ha and[N II]6584-A lines to be #30 per cent (Table 1). The flux fromthe broader Ha component (distinct from [N II]) is very poorlyconstrained.

Note that although the narrower Ha line is unresolved (FWHMless than ,200 km s¹1 at a resolving power of 1428), there is

A spectroscopic study of IRAS F10214 þ 4724 325

q 1998 RAS, MNRAS 298, 321–331

Figure 5. H-band spectrum of F10214 þ 4724. The position of the [O III]doublet is marked, as are the predicted positions of other emission lines atz ¼ 2:286.

Figure 6. K-band spectrum of F10214 þ 4724. The position of the narrowerHa and [N II] 6584-A lines are marked assuming z ¼ 2:286, as is thepredicted position of the [N II] 6548-A line. Note the clear broad base tothese lines.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

marginal evidence that the [N II] 6584-A line is resolved. We foundthat at least some of the apparent excess on the [N II] line, if real,may be attributable to the broader [N II] component. If we assumethat the satellite lines contribute 60 per cent of this flux, we obtain abroader Ha flux of 75 per cent of the total Ha. Note that the modeldoes not imply the narrowest lines are Gaussian, but rather than theK-band spectrum must have at least one unresolved and at least oneresolved component. This is exactly as expected for a hybridstarburst/AGN system.

To summarize, we find three distinct kinematic components inour new spectra of F10214 þ 4724: first, a ,1000 km s¹1 compo-nent, present in Ha and in the UV lines which comes from theSeyfert 2 narrow-line region (NLR), as will be discussed further inSection 4; second, a < 200 km s¹1 component, present in Ha, whichwe chose to associate with starburst activity, as will be discussed inSection 5; and, third, a 4000 km s¹1 C III]1909 line which originatesin a hidden Seyfert 1, or quasar, nucleus, as has been discussed byGoodrich et al. (1996).

4 T H E S E Y F E RT 2 N A R ROW- L I N E R E G I O N

4.1 Resonant scattering of Lyman series lines

There is strong evidence for resonant scattering of the Lya emissionline: the profile of Lya is double-peaked, and the peaks aresymmetrically displaced about the predicted position of Lya atz ¼ 2:286 with equal flux (within the errors; Table 1). If the Lya-emitting region is free of dust, then the velocity shift v of either Lya

peak from 1215.7 A is given by

v ¼ 195NH

1023 m¹2

� �1=3 T

104 K

� �1=6

km s¹1 ð1Þ

(Neufeld & McKee 1988), where NH is the total atomic hydrogencolumn density to the source of the Lya photons, and T is the gastemperature. The measured velocity separation of the peaks is2 × ,560 km s¹1, yielding NH ¼ 2:5 × 1025 m¹2 for T ¼ 104 K.

The total line-centre optical depth to the source of the Lya

photons is tLya , NH × kLya, where kLya is the absorption cross-section to the line centre of Lya. In the general case of photonabsorption promoting an electron upwards in energy from level i to

level j, the line-centre cross-section kl is given by

kl ¼1

4pe0

����p

pe2fij

mcDnD; ð2Þ

where fij is the oscillator strength for the transition, m and e are theelectronic mass and charge, c is the speed of light, and e0 is thevacuum permittivity; DnD ¼ ð2kT=MAc2Þ0:5 × n0 is the Dopplerwidth of the line with n0 the central frequency, k is the Boltzmannconstant, and MA is the atomic mass. In the case of the Lya line,taking T ¼ 104 K, these equations yield kLya ¼ 5:8 × 10¹18 m2, andthus tLya ¼ 1:5 × 108.

Resonant scattering is extremely efficient at removing Lyb photonseven at moderate optical depths (e.g. Netzer 1975; Osterbrock 1989),in agreement with the null detection of Lyb in F10214 þ 4724. Thestrong influence of resonant scattering on the observed properties ofthe Lyman series lines is a major difference between the spectroscopicproperties of F10214 þ 4724 and NGC 1068. UV spectroscopy of thelatter (Kriss et al. 1992b) shows a strong single-peaked Lya line aswell as Lyb; the observed Lyb/Lya flux ratio of 0.12 impliestLya & 10.

Is this neutral material within the NLR, or external, such as adamped Lya system associated with the host galaxy? The QSOproximity effect (e.g. Giallongo et al. 1996) suggests that neutralmaterial is often destroyed within the ionization cone. If the NH isassociated with the host, it need not cause damped absorption inF10214 þ 4724 itself (countering the most obvious criticism),because the geometry would allow scattering into the line ofsight. This neutral hydrogen could be intrinsic to the host, orcould be the result of accretion of gas-rich dwarfs; neutral systemswith column densities $1025 m¹2 are known to undergo a strongcosmological evolution which dominates the evolution in Qg (Wolfeet al. 1995).

Can the NH gas-to-dust ratio, derived from the Lya strength, beused to resolve this? Lya photons are conserved in resonantscattering, but the Lya strength can be suppressed by dust absorp-tion between scatterings. The predicted case B Lya flux from ourbroader Ha flux implies a suppression factor of < 102, the limitassuming no [N II] 6548, 6584. In the damped Lya models ofCharlot & Fall (1991), a neutral screen with velocity dispersion10 km s¹1 and column density as derived above yields a Lya

suppression of ,104, but if the gas-to-dust ratio is ,4 times lowerthan the values typically inferred from the reddening of quasarswith damped systems, this can be reconciled with our observed Lya.The Lya flux therefore cannot exclude the possibility of a dampedsystem associated with the host. Alternatively, the same modelspredict a suppression ,10¹1 for sources distributed throughout theNH (for a velocity dispersion of 1000 km s¹1 the suppressionbecomes only ,0:75), and the authors argue that geometricalfactors could reduce this to unity.

There are difficulties with resonant scattering within the NLR:the clouds in the nearer ionization cone suffer resonant scattering,but we nevertheless ought to have an unobscured view of the cloudsin the more distant ionization cone. This contradicts, for example,the observed lack of Lyb, but this could be explained by differentialmagnification of the nearer ionization cone, or selective obscurationof the more distant ionization cone, e.g., by the host galaxy.

We nevertheless favour models with at least some of the neutralcolumn in the rear of the narrow-line clouds themselves, since thephotoionization models below (Section 4.2) favour ionization-bounded clouds. Higher resolution K-band spectroscopy mayresolve the AGN Ha from the satellite [N II] lines, and the inferredLya suppression may exclude the damped-system model. Spatially

326 S. Serjeant et al.

q 1998 RAS, MNRAS 298, 321–331

Figure 7. Fit to the K-band spectrum of F10214 þ 4724 discussed in thetext.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

resolved spectroscopy of F10214 þ 4724 (while subject to theuncertainties in the projection to the source plane) may provide afurther confirmation: diffuse Lya could be used to trace anyextended neutral hydrogen (Villar-Martin, Binette & Fosbury1996), modulo the uncertainties in the projection to the sourceplane. The detection of non-resonance lines spatially coincidentwith any extended Lya would rule out resonance scattering‘mirrors’ and identify such emission with the AGN NLR orstarburst knots.

4.2 Photoionization models

If we exclude the resonantly scattered Lya, the marginally resolveddoublet N V1240, 1243, and the blended emission lines (labelled inTable 1), then we find no correlation of ionization potential withemission linewidth. As a result, we may attempt to model theemission-line spectrum with constant-density, single-slab photo-ionization models, at least to first order.

We modelled the Seyfert 2 lines with the photoionization codeCLOUDY (version 84.12, Ferland 1993), assuming for simplicity thatthe rest frame UV emission lines in F10214 þ 4724 suffer zeroextinction (Section 4.4). Unless otherwise stated, we used an AGNionizing continuum as characterized by Matthews & Ferland (1987)with a break at 10 mm (see Ferland 1993). We considered hydrogendensities 8 # log10 n # 16, where n is the total hydrogen columndensity (i.e., molecular, neutral and ionized) in units of m¹3, andionization parameters ¹4 # log10 U # 0, where U is the dimen-sionless ratio of incident ionizing photons to hydrogen density, i.e.,U ¼ FðHÞðncÞ. Here F is the surface flux of ionizing photons inm¹2 s¹1, and c is the speed of light. The calculation was stopped at acolumn density of 1027 m¹2, or at a temperature of 4000 K, sincebelow this temperature the emission-line flux is negligible. We usedthe default solar metallicity (see Ferland 1993), and CLOUDYassumesa plane-parallel geometry.

The He II:Lya:O VI:C IV ratios are sensitive to the UV–soft X-raycontinuum (e.g. Krolik & Kallman 1988), so are sensitive to thepresence or absence of an active nucleus. A blackbody ionizingcontinuum at 40 000 K, resembling the continuum from the hottestOB stars, has severe difficulty in producing O VI1032, 1037 A orMg VI 1086 A, due to the extremely high ionization potentials ofO4þ and Mg4þ (0.11 and 0.14 keV respectively). The AGN ionizingcontinuum defined above had no difficulty in reproducing theselines. We are therefore confident that the ,1000 km s¹1 emission-line region is predominantly photoionized by the active nucleus.

None of the single-slab models provided an adequate fit to all therest frame UV emission lines. We found that similar, but notidentical, conditions (log10 n , 10, log10 U , ¹1:5) were impliedby emission lines [Ne IV] 1602, He II 1640, N IV] 1486, N III] 1750,Ne V 3426, Ne III 3869. For example, the largest [Ne IV] 2424/He II

1640 ratio in our models was with a density and ionizationparameter of log10 ¼ 10, log10 U ¼ ¹1:4. This underpredicts theO VI and Mg VI lines, as well as the N lines. However, the N III] 1486,1750 and N V 1240 ratios are all consistent with log10 U , ¹1:25;this self-consistency of the N lines suggests that the strength of theN 1240-A line may be due mainly to an abundance effect, as alsoargued in the IRAS-detected Seyfert 2 galaxy NGC 3393 (Diaz et al.1988), rather than due to differential magnification (Broadhurst &Lehar 1995). If so, this would further reduce the proportion of Ha inthe broader component of Fig. 7.

This model reproduced the C IV:C III] ratio, but over-predictedtheir strength by a factor of ,5. A low carbon abundance (,0:5×solar) might resolve the anomaly, perhaps caused by depletion on to

dust grains. Much more problematic is the lack of an [O III] 1665-Aemission line, which in these conditions should have a flux ,10 percent of He II 1640 A. There does not appear to be any simplesolution; we note, however, that an identical problem was found byKriss et al. (1992b) in NGC 1068, who argued that shock-heated gaswould also produce the [O III] 1665 line.

Finally, we can use our limits on the density and ionizationparameter to derive the sizes and masses of the NLR clouds. Theclouds are ionization-bounded in the conditions we derived fromthe emission-line ratios, as is also conventionally assumed for AGNNLRs, and unlike, for example, the matter-bounded extendedemission-line region models of Wilson, Walker & Thornley(1997). This supports our model for the resonant scattering (Section4.1), and is also consistent with the presence of [O I] in the Keckspectrum (Soifer et al. 1995). Assuming that the absorbing columndoes indeed occur within the NLR clouds (assumed to be constant-density for simplicity), we obtain a characteristic size-scale of

S ¼ NH=nH ¼ 2:5 × 10¹260:5 pc; ð3Þ

where the errors represent the acceptable range in the models, ratherthan, for example, Gaussian noise. The mean mass of the NLRclouds is given by

M ¼ u × nH × S3 ¼ 1 × 10¹261 M(: ð4Þ

Unfortunately, it is not possible to derive a covering factor ofvolume filling factor without spatially resolved spectroscopy.Briefly, N clouds with covering factor C at mean distance D parsecsfrom a quasar with luminosity L are indistinguishable from N, C=4,2D, 4L. We can, however, estimate the number of narrow-lineclouds, and the total cloud mass contained in the NLR. UsingCLOUDY, we find the emissivity of a single cloud of density1010:560:5 m¹3 with ionization parameter 10¹1:2560:25; e.g., for theHe II 1640 line e ¼ 10¹2:460:7 W m¹2. For a magnification factor of10M10, our observed line flux implies the number of narrow-lineclouds is

N ¼ ð10¹18 W m¹2 × 4pD2LÞ=ð4peS2 × 10M10Þ

¼ 1 × 10761:4M¹110 h¹2

50 ;

ð5Þ

where DL is the luminosity distance to F10214 þ 4724, given(assuming Q0 ¼ 1, L ¼ 0, H0 ¼ 50h50 km s¹1 Mpc¹1) by

DL ¼ 0:04ch¹150 ð1 ¹ ð1 þ 2:286Þ¹0:5Þð1 þ 2:286Þ: ð6Þ

(We also note that variations in S correlate with variations in thecloud emissivity, so the ‘errors’ do not simply add.) The inferredtotal mass of ionized gas within the narrow-line region is consistentwith estimates in local AGN (e.g. Osterbrock 1993):

NM ¼ 10¹184pD2LunHS3

=ð4pS2e10M10Þ

¼ NH × 5 × 101260:7m2M¹110 h¹2

50 kg

¼ 2 × 10560:7M¹110 h¹2

50 M(:

ð7Þ

Our choice of He II, while free of metallicity effects, assumesnegligible extinction (Section 4.4), which will be testable withhigher quality infrared spectra. Note that if spatially resolvedspectroscopy detects the emission-line counterimage, it may bepossible to determine M10 independently of spatial structure in theIRAS galaxy. As a corollary to the covering factor calculation, onemay then also obtain a robust estimate of the luminosity of theobscured quasar (Goodrich et al. 1996).

4.3 Optically thick O VI emission

The resonant O VI 1034, N V 1240 and C IV 1549 lines are doublets in

A spectroscopic study of IRAS F10214 þ 4724 327

q 1998 RAS, MNRAS 298, 321–331

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

the isoelectronic lithium sequence. If such lines emanate from aregion which is optically thin to resonant scattering, then theirdoublet line ratios are given by the ratio of the statistical weights ofthe upper levels, i.e., 2:1 with the lower wavelength line of thedoublet the brighter. However, in optically thick regions, photontrapping and collision de-excitation can cause the levels to therma-lize so that the flux ratio approaches ,1:1; this thermalizationoccurs only at higher values of optical depth and electron density ne,namely when net0 * 1022 m¹3 (Hamann et al. 1995). A doubletratio of 1:1 therefore implies extreme densities and/or opticaldepths.

The O VI 1031.9, 1037.6 doublet in F10214 þ 4724 is clearlyresolved (Fig. 4), and its flux ratio 1:1 suggests line thermalization.We first consider whether this ratio is strongly influenced byblanketing by the Lya forest lines – both a simple empirical testand a brief analytical argument suggest that it is not. The testinvolved fitting low-order polynomials to the UV continuum long-ward of Lya, and extrapolating these fits to shorter wavelengths toestimate the level of line blanketing. We then superimposed a model2:1 doublet at several positions on this absorption spectrum, andfound it impossible to obtain an apparent 1:1 ratio. This is in goodagreement with the following analytical argument. The number ofLya forest lines per unit redshift per unit restframe equivalent widthis well fitted by an expression of the form

d2NdzdW

¼A0

W¬ð1 þ zÞg exp ¹

WW¬

� �; ð8Þ

where g . 1:89, W¬ . 0:27 A and A0 . 10 (e.g. Bechtold 1994).By integrating this over the redshift range 1.770 to 1.792 (i.e., in theneighbourhood of the doublet and exactly encompassing the widthof either line of the doublet), and further integrating dN=dW , it canbe shown that a maximum of two Lya forest lines are expected to lieon the position of one O VI line. For a 2:1 doublet to be suppressed to1:1, the line blanketing over the shorter wavelength line must beapproximately a factor of 2 (in addition to the mean blanketing); theprobability of this occurring is negligibly small, a conclusion whichis insensitive to the uncertainties in g, W¬ and A0.

This seems to imply that the O VI-emitting region inF10214 þ 4724 is both optically thick (tO vi q 1) and denseenough for the doublet to have become thermalized, i.e.,netO vi * 1022 m¹3. We will next show that the optical depth tO vi

is unlikely to exceed 106, resulting in apparent BLR-like conditionsfor the narrow O VI-emitting region.

The latest release of CLOUDY (version 90.03a, Ferland 1996)incorporates the O VI 1032, 1037-A doublet lines separately, sowe used this code to explore the doublet ratios and O VI opticaldepths for a wide range of physical conditions. We used the sameAGN ionizing continuum, metallicity and stopping conditions asabove. The results are shown in Fig. 8. We find that the 1:1 ratio,also where the net0 * 1022 m¹3 limit applies, is possible only withhydrogen densities in excess of nH . 1017 m¹3. This conclusionwas found to apply equally to NH ¼ 1026 m¹2 column and/orZ ¼ 10 Z( metallicity gas. Such a high value of nH is characteristicof the BLRs of AGN (e.g. Ferland et al. 1992) rather than the lowervalues (ne , 1011 m¹3) found in classical NLRs.

Such a radical conclusion can be avoided, albeit with some fine-tuning, by considering the Lyb resonant absorption. This absorptionis likely to be saturated, so the O VI doublet may lie inside itsdamping wings. Using the NH value derived above, we obtainan optical depth of about 0.3 for the 1032-A line, and 0.09 for the1038-A line (i.e., a factor ,3:6). To obtain a 1:1 ratio from 2:1, oneneeds an optical depth of around 1 to the 1032-A line, close to the

estimate from the observed NH. This explanation avoids the highdensities, but is sensitive to the assumed neutral column, andfurthermore takes no account of geometrical effects (Villar-Martin et al. 1996).

Inspection of the UV spectrum of NGC 1068 (Kriss et al. 1992b)suggests that the O VI doublet is again in a ,1 : 1 ratio. However,there is no evidence of resonant scattering in this object.

4.4 Reddening of the Seyfert 2 nucleus

We can place limits on the extinction in the observed UV from theHe II 1640 A/He II 1086 A ratio, which in the absence of reddening ispredicted to be 7 (e.g. Seaton 1978). Although the signal-to-noiseratio shortward of Lya in Fig. 4 is poor, and despite possibleblanketing by Lya forest lines, the He II 1086-A line is marginallydetected at (,7 6 50 per cent)¹1 times the flux of the 1640-Acounterpart. If real, this suggests an extinction of between zero and0.75 mag around 1200 A in the rest frame of F10214 þ 4724 (i.e.,AV < 0:2), or 2600 A in that of the lens (i.e., AV < 0:4). We can alsoobtain a rough extinction estimate from the Ha:He II 1640 A ratio,predicted to be about 2.3 in our solar-metallicity models above. Wefind (Ha þ [N II]):He II ¼ 3:8, consistent with an Ha:N II ratioresembling, e.g., NGC 1068 (e.g. Bland-Hawthorn, Sokolowski &Cecil 1991) and zero extinction.

Comparison of the H-band detection of the He II 4686-A line (e.g.Soifer et al. 1995) with the 1640-A line suggests extinctionsAV , 1:3. However, our photoionization models above found theFe II 4658-A and/or Ar IV 4720-A lines can contribute up to 3 timesthe flux of the He II 4686-A line in densities of 1010 –1012 m¹3, notunreasonable for AGN NRLs. Moreover, the Soifer et al. H-bandspectrum has very low spectral resolution, and the He II 4686-A lineflux must be treated with some caution.

This low reddening in the rest frame UV is in marked contrast tothe high (AV $ 5) rest frame optical extinction derived from Balmerdecrements (e.g. Soifer et al. 1995), which led several authors toinfer that at least two physically distinct regions contribute to theobserved emission-line spectrum (e.g. Elston et al. 1994; Soifer etal. 1995). The limit on the Balmer decrement of Ha:Hb $ 20reported by Elston et al. (1994) is at least in part resolved by our

328 S. Serjeant et al.

q 1998 RAS, MNRAS 298, 321–331

Figure 8. Results of the photoionization model discussed in the text. Thecontours show the product of the np-weighted average electron density andO VI 1032 þ 1037-A optical depth, ne × tO vi, as functions of ionizationparameter and density. (Almost identical results were obtained replacing thisne with the innermost zone electron density, the outermost, or the totalhydrogen density.) The contours are spaced logarithmically in steps of 0.5.Only where ne × tO vi * 1022 m¹3 does the O VI doublet form a 1:1 ratio.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

K-band spectrum. The narrower Ha component contributes ,25per cent of the total Ha flux; furthermore, our total Ha flux is ,25per cent lower than that reported by Elston et al., although this iswithin their stated photometric errors. However, this still yieldsAV > 4 in the frame of F10214 þ 4724, or AV > 6 in that of the lens.(Any < 200 km s¹1 Hb leads to an even higher AV.) Alternatively, ifwe assume that the Iwamuro et al. (1995) Hb detection is correct,the data are consistent with zero extinction. Further near-infraredspectroscopy is required to resolve this issue. We conclude that asyet there is no conclusive evidence for any significant reddening ofthe Seyfert 2 nucleus.

5 S TA R B U R S T P R O P E RT I E S

Our limit on the width of the narrowest Ha component inF10214 þ 4724 is suggestive of a star-forming region, as haspreviously been suggested by Kroker et al. (1996). Moreover, thefluxes of the [O I] 6300-A and [S II] 6724-A lines (Soifer et al. 1995)relative to the narrowest Ha component are similar to both thoseseen in spectra of local starburst galaxies (e.g. De Robertis & Shaw1988) and in high-luminosity IRAS galaxies (e.g. Leech et al. 1989).We can obtain an estimate of the star formation rate R from thenarrower Ha line flux S, using the following expression adaptedfrom Kennicutt (1983):

Rðz ¼ 2:286Þ ¼SðHaÞh¹2

50

3 × 10¹21 W m¹2 M( yr¹1: ð9Þ

We assume a Hubble constant of H0 ¼ 50 h50 km s¹1 Mpc¹1 withQ0 ¼ 1, QL ¼ 0. If we assume that the starburst region undergoes anet gravitational magnification of M & 10 (Downes et al. 1995;Graham & Liu 1995; Green & Rowan-Robinson 1996), and thataperture corrections are negligible, we obtain a star formation rateof * 20h¹2

50 M( per year. This corresponds to a starburst luminosityof * 2:6 × 1011h¹2

50 L( (e.g. Moorwood 1996), i.e., ,0:5 per cent ofthe magnification-corrected bolometric power output.

Reddening by intervening dust would further increase thisestimate of the star formation rate. Radiative transfer models ofF10214 þ 4724 (Green & Rowan-Robinson 1996) suggest a muchlarger starburst bolometric fraction of ,0:2–0.4, with a highstarburst UVoptical depth of tUV , 800, i.e., an AV , 160. Assum-ing that this dust is well mixed with the Ha-emitting gas (e.g.Thronson et al. 1990), the extinction and star formation rate is quiteconsistent with our observed narrow Ha flux, which for this AV

yields a starburst bolometric fraction of ,0:5. In both NGC 1068(Green & Rown-Robinson 1996) and F10214 þ 4724, the starburstand active nuclei appear to make comparable contributions to thebolometric power output.

6 C O N C L U D I N G R E M A R K S

The presence of the optically thick O VI 1032, 1037-A doubletappears at face value to imply extremely high densities for narrowemission-line gas, nH * 1017 m¹3. However, we argue that it ismore easily attributable to the Lyb damping wings of resonantscattering material, probably within the AGN NLR. Differentialmagnification appears to play a significant role in the emission-linespectrum of F10214 þ 4724, in which we identify three distinctkinematic components: quasar broad lines (,4000 km s¹1), Seyfert2 narrow lines (,1000 km s¹1) and a starburst component (&200km s¹1). The density, ionization parameter, number and total massof Seyfert 2 narrow-line clouds all resemble local Seyferts, and ourinterpretation of the resonant scattering agrees with that of Villar-Martin et al. (1996) for high-redshift radio galaxies. The flux from

the narrowest Ha component is in excellent agreement withradiative transfer models which comprise similar quasar andstarburst bolometric contributions.

AC K N OW L E D G M E N T S

We thank Carlos Martin for assisting the observations. The WHT isoperated on the island of La Palma by the Royal GreenwichObservatory in the Spanish Observatorio del Roque de los Mucha-chos of the Instituto de Astrofisica de Canarias. We thank SteveEales for performing the GASP astrometry of the F10214 þ 4724field, and for (unwittingly) contributing observing time to theproject. We also thank Tony Lynas-Gray, Geoff Smith and SteveWarren for useful discussions.

REFERENCES

Adams T., 1972, MNRAS, 174, 439Antonucci R., Hurt T., Miller J., 1994, ApJ, 430, 210Barvainis R., Tacconi L., Antonucci R., Allion D., Coleman P., 1994, Nat,

371, 586Bechtold J., 1994, ApJS, 91, 1Blandford R. D., Kochanek C. S., 1987, ApJ, 321, 658Bland-Hawthorn J., Sokolowski J., Cecil G., 1991, ApJ, 375, 78Broadhurst T., Lehar J., 1995, ApJ, 450, L41Bruzual A. G., Charlot S., 1993, ApJ, 405, 538Charlot S., Fall S. M., 1991, ApJ, 378, 471Clements D., van der Werf P., Krabbe A., Blietz M., Genzel R., Ward M.,

1993, MNRAS, 262, 23P

Close L. M., Hall, P. B., Liu C. T., Hege E. K., 1995, ApJ, 452, L9Davidson K., Netzer H., 1979, Rev. Mod. Phys., 51, 715De Robertis M. M., Shaw R. A., 1988, ApJ, 329, 629Diaz A. I., Prieto M. A., Wamsteker W., 1988, A&A, 195, 53Downes D., Radford S. J. E., Greve A., Thum C., Solomon P. M., Wink J. E.,

1992, ApJ, 398, L25Downes D., Solomon P. M., Radford S. J. E., 1995, ApJ, 453, 65Dunlop J. S., Hughes D. H., Rawlings S., Eales S. A., Ward M. J., 1994, Nat,

370, 347Eisenhardt P. R., Armus L., Hogg D. W., Soifer B. T., Neugebauer G.,

Werner M. W., 1996, ApJ, 461, 72Elston R., McCarthy P. J., Eisenhardt P., Dickinson M., Spinrad H., Januzzi

B. T., Maloney P., 1994, AJ, 107, 910Ferland G. J., 1993, University of Kentucky Department of Physics and

Astronomy Internal ReportFerland G. J., 1996, Hazy, a Brief Introduction to Cloudy. University of

Kentucky Department of Physics and Astronomy Internal ReportFerland G. S., Peterson B. M., Horne K., Welsh, W. F., Nahar S. N., 1992,

ApJ, 387, 95Giallongo E., Cristiani S., D’Odorico S., Fontana A., Savaglio S., 1996, ApJ,

466, 46Goodrich R., Miller J., Martel A., Cohen M., Tran H., 1996, ApJ, 456, L9Graham J. R., Liu M. C., 1995, ApJ, 449, L29Green S. M., Rowan-Robinson M., 1996, MNRAS, 279, 884Hamann F., Shields J. C., Ferland G. J., Korista K. T., 1995, ApJ, 454, 688Hines D. C., Schmidt G. D., Cutri R. M., Low F. J., 1995, ApJ, 450, L1Hughes D. H., Dunlop J. S., Rawlings S., 1997, MNRAS, 289, 766Iwamuro F., Maihara T., Tsukamoto H., Oya S., Hall D. B., Cowie L. L.,

1995, PASJ, 47, 265Jannuzi B. T., Elston R., Schmidt G. D., Smith P. S., Stockman H. S., 1994,

ApJ, 429, L49Kennicutt R. C., Jr., 1983, ApJ, 272, 54Kennicutt R. C., Jr., Tamblyn P., Congdon C. E., 1994, ApJ, 435, 22Kriss G. A. et al., 1992a, ApJ, 392, 485Kriss G. A. et al., 1992b, ApJ, 394, L37Kroker H., Genzel R., Krabbe A., Tacconi-Garman L. E., Tecza M.,

Thatte N., 1996, ApJ, 463, L55

A spectroscopic study of IRAS F10214 þ 4724 329

q 1998 RAS, MNRAS 298, 321–331

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

Krolik J. H., Kallman T. R., 1988, ApJ, 324, 714Lacy M., Rawlings S., 1994, MNRAS, 270, 431Laor A. et al., 1994, ApJ, 420, 110Lawrence A. et al., 1993, MNRAS, 260, 268Lawrence A., Rigopoulou D., Rowan-Robinson M., McMahon R. G.,

Broadhurst T., Lonsdale C. J., 1994, MNRAS, 266, L41Leech K. J., Penston M. V., Terlevich R., Lawrence A., Rowan-Robinson M.,

Crawford C., 1989, MNRAS, 240, 349Matthews W. G., Ferland G. J., 1987, ApJ, 323, 456Matthews K. et al., 1994, ApJ, 420, L13Mendoza C., 1983, in Planetary nebulae: Proc. Symp. London. Reidel,

Dordrecht, p. 143Moorwood A. F. M., 1996, Space Sci. Rev., 77, 303Mountain C. M., Robertson D. J., Lee T. J., Wade R., 1990, in Crawford

D. L., ed., Instrumentation in Astronomy VII. Proc. SPIE, 1235, 25Nelson C. H., Whittle M., 1996, ApJ, 465, 96Netzer H., 1975, MNRAS, 171, 395Neufeld D. A., McKee C. F., 1988, ApJ, 331, L87Osterbrock D. E., 1989, Astrophysics of Gaseous Nebulae and Active

Galactic Nuclei. University Science Publications, Mill Valley, CA, USAOsterbrock D. E., 1993, ApJ, 404, 551Radford S. J. E., Brown R. L., Vanden Bout P. A., 1993, A&A, 271, L21Ramsay S. K., Mountain C. M., Geballe T. R., 1992, MNRAS, 259, 751Rowan-Robinson M. et al., 1991, Nat, 351, 719Rowan-Robinson M. et al., 1993, MNRAS, 261, 513Scoville N. Z., Yun M. S., Brown R. L., Vanden Bout P. A., 1995, ApJ, 449,

L109Seaton M. J., 1978, MNRAS, 185, 5P

Serjeant S., Lacy M., Rawlings S., King L. J., Clements D. L., 1995,MNRAS, 276, L31

Snidjers M. A. J., Netzer H., Boksenberg A., 1986, MNRAS, 222, 549Soifer B. T., Neugebauer G., Matthews K., Lawrence C., Mazzarella J.,

1992, ApJ, 399, L55Soifer B. T., Cohen L., Armus K., Matthews G., Neugebauer G., Oke J. B.,

1995, ApJ, 443, L65Solomon P. M., Downes D., Radford S. J. E., 1992, ApJ, 398, L29Thronson H. A., Majewski S., Descartes L., Hereld M., 1990, ApJ, 364,

456Trentham N., 1995, MNRAS, 277, 616Villar-Martin M., Binette L., Fosbury R. A. E., 1996, A&A, 312, 751Warren S., Hewett P., Osmer P., 1994, ApJ, 421, 412Wills B. J., Netzer H., Wills D., 1980, ApJ, 242, L1Wilson C. D., Walker C. E., Thornley M. D., 1997, ApJ, 482, 131Wolfe A. M., Lanzetta K. M., Foltz C. B., Chaffee F. H., 1995, ApJ, 454,

698

A P P E N D I X A : C O M PA N I O N G A L A X I E S

Fig. A1 shows the spectrum of a galaxy ,19 arcsec south-west ofthe IRAS galaxy, which fortuitously lay on the slit in our WHTobservations. The feature at around 5510 A is probably a low-energy cosmic ray event, but the emission lines at 7155 A and(marginally) at 5323 A appear to be real, since they occur in thespectra on both nights. Cross-correlating this spectrum with oldgalaxy models (a 1-Gyr burst aged by 6 × 108 yr to 1 Gyr) fromBruzual & Charlot (1993) yields two redshift estimates, 0.428 and0.476 (both 6 , 5 per cent). We prefer the former, which identifiesthe weak emission features as [O II]3727 and [O III]5007. Close et al.(1995) report a similar redshift, 0:429 6 0:002 for a further galaxy,23 arcsec south-west of the IRAS galaxy. The UV upturn in Fig.A1 may indicate a recent starburst, as expected perhaps if thegalaxies are tidally interacting. However, there could be a slightsystematic error in the redshift of the Close et al. companion. Theemission-line wavelengths in their spectrum of F10214 þ 4724 areoffset by a factor of ,0:99 compared to other published spectra (e.g.Rowan-Robinson et al. 1993; Goodrich et al. 1996), and the

residuals from the 5577-A sky feature also appear to be offset bya similar amount.

The tentative lens redshift of z . 0:90 (Serjeant et al. 1995) isconsistent with both its velocity dispersion (e.g. Broadhurst &Lehar 1995) and surface brightness profile (Eisenhardt et al.1996). Further support for this lens redshift comes from Fig. A2,in which the spectra of sources 2 and 3 (Serjeant et al. 1995) aresummed. Source 2 is the lensing galaxy, and source 3 is thecompanion ,3:3 arcsec north-east of the IRAS galaxy, in thenotation of Matthews et al. (1994). If the galaxies are indeedassociated, as suggested by their similar spectral energy distribu-tions and tentative 4000-A breaks (Serjeant et al. 1995), as well asby hints of tidal interaction in HST imaging (Eisenhardt et al. 1996),then the summed spectrum should show a ,

���2

pimprovement in

signal-to-noise ratio, which is indeed the case in Fig. A2.Several groups have also noted the slight over-density of galaxies

in the vicinity of F10214 þ 4724. The discovery of a Mg II absorber

330 S. Serjeant et al.

q 1998 RAS, MNRAS 298, 321–331

Figure A1. Serendipitous galaxy ,19 arcsec south-west of F10214 þ 4724.The flux has units 10¹20 W m¹2 A¹1, and the spectrum is extracted with afull-width zero-intensity aperture.

Figure A2. The sum of the spectra of sources 2 and 3 (in the notation ofMatthews et al. 1994). These spectra were extracted with 3-pixel (,1:1arcsec) apertures, differing slightly from the method of Serjeant et al. 1995,though with very little quantitative difference. Source 2 is corrected forcontamination by the IRAS galaxy by subtracting a spectrum ofF10214 þ 4724 scaled by the strength of the C III]1909 line. The sum hasbeen smoothed with a 5-pixel (,14-A) boxcar. The flux has units 10¹20 Wm¹2 A¹1.

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from

in the Keck spectrum at z ¼ 1:316 (Goodrich et al. 1996), and thepossibly associated pair at z . 0:4, implies that several physicallyindependent systems, at a variety of redshifts, contribute to the over-density in the field.

A P P E N D I X B : T H E M Y S T E RY L I N E AT 2 0 6 7 A

In this appendix we summarize our attempts to identify the line at2067 A in the rest frame spectrum of F10214 þ 4724. We havemade an extensive literature search, and investigated syntheticspectra generated by the CLOUDY photoionization code over wideranges of ionization parameter and abundances (version 84.12,Ferland 1993; see also Section 4.2). This led to only two possibleidentifications for this line. The first of these is the resonant boronB III 2068 doublet, the next resonance line in the lithium isoelec-tronic sequence below the bright O VI, N V and C IV lines. The tinycosmic abundance of boron means that this line is predicted to beimmeasurably faint, and we therefore reject this possibility. Thesecond possibility is the forbidden [Na V] 2067.9/2069.8 doublet(Mendoza 1983) – a line which is not found in the CLOUDY version84.12 output.

To determine whether the [Na V] 2067.9/2069.8 doublet is aplausible identification for the mystery line, we have comparedits predicted strength with that of [Ne IV] 2422.5/2425.1. These twopairs of collisionally excited lines arise from electronic transitionswithin adjacent ions in the nitrogen isoelectronic sequence. In the

low-density limit the ratio of the cooling rates in these doublets isgiven by

L2069

L2424¼

nNa v

nNe iv×

e¹x2069=kT

e¹x2424=kT ×Q2069

Q2424×

24242069

ð10Þ

(Osterbrock 1989), where xl is the excitation potential of the upperenergy level(s), and Ql is the effective collision strength (seeMendoza 1983).

Adopting T ¼ 104 K and solar abundances, and assuming thatthe fraction of Na in the Na4þ state is similar to the fraction of Ne inthe Ne3þ state, the value of this ratio is <0:005. The measured ratiois <0:23, so at first sight the putative [Na V] 2067.9/2069.8 doubletwould need to be anomalously strong. However, at higher densities,collisional de-excitation may become important. The critical den-sities for collisional de-excitation ncrit are ,1011 and ,1012 m¹3 forthe lower wavelength lines in the Ne and Na doublets respectively(in each case the higher wavelength doublet line has an order ofmagnitude lower ncrit than its partner). Given the weakness of the[O II]3727 line, with ncrit , 1010 m¹3 in F10214 þ 4724, and thestrengths of the [O III]4959/5007 line and forbidden neon lines(with ncrit * 1012 m¹3) (Soifer et al. 1995), it is certainly feasiblethat collisional de-excitation can account for the strength of theputative [Na V] doublet with respect to the [Ne IV] doublet.

We conclude that the [Na V]2067.9/2069.8 doublet is a probableidentification for the mystery line in the spectrum of F10214 þ 4724and, by analogy, in NGC 1068 and other active galaxies.

A spectroscopic study of IRAS F10214 þ 4724 331

q 1998 RAS, MNRAS 298, 321–331

at Open U

niversity Library (PE

R) on A

pril 13, 2016http://m

nras.oxfordjournals.org/D

ownloaded from